RCA-NET: Image Recovery Network with Channel Attention Group for Image Dehazing

In this paper, we propose an image dehazing network with a channel attention model. Most existing methods try to resolve the dehazing problem through an atmospheric transmission model, but always fail to get promising results since the real-world physical imaging system is of high complexity. Therefore, we propose recovering a fog-free image from its foggy image using an end-to-end pipeline which can produce more realistic results. We apply a channel-wise attention model into our network and also employ the perceptual loss for supervision. Experimental results indicate that our method performs better than several state-of-the-art algorithms.